Certain flash memory may be made up of arrays of NAND strings of transistors organized into large blocks, wherein erase operations in NAND flash may typically occur at the block level. Thus, a given erase operation (e.g., “erase”) may apply to an entire block under conventional approaches. As systems increase in scale to include more transistors, block size may continue to increase exponentially. As a result, a substantial number of strings may be over-erased at the end of a given erase operation (e.g., with the end of the erase operation being dictated by the slowest string to erase). Furthermore, the inability to erase at higher granularity may lead to an increased number of write operations in devices such as solid state disks (SSDs).